Collaborative Modelling Approaches

There are some stages in the design and/or use of models that involve several persons. The term ‘collaborative’, or ‘participatory’, modelling refers to such group model building situations where model designers and/or users are actively involved (Eden et al., 1996).

6.2.1. Collaborative Modelling Background

Collaborative modelling approaches stem from system dynamics, and usually refer to integrated ecological modelling. The idea is that a system’s point of view is helpful to

‘lift’ observers to the system’s level and to create a holistic view. Several methods have been developed to support such an idea, ranging from problem structuring and the system’s conceptualization (e.g. cognitive mapping) to the construction of computer models simulating the behaviours of stakeholders interacting within a complex socio-ecosystem.

Goals and objectives

Model conceptualization

Model implementation Previous knowledge

and experimental data

Scenario runs

Decision support

Research and understanding Model use

Problem identification

Model analysis Sensitivity analysis Calibration

Verification Validation

Source: adapted from Voinov and Bouman, 2004.

Figure 6.1 An example of the sequentiality of the modelling process.

In ecology, hard and soft systems are often referred to (Röling and Wagmaker, 1998). Hard systems are treated as if the systems really exist. The systems’

boundaries and goals are assumed to be given. Analysis and problem solving focus on goal seeking and the best technical means to reach a goal. The crucial assumption is that system goals are not given but contested, and that system boundaries are negotiated. The soft systems methodology emphasizes a group of actors who are faced with a shared problem to engage in a collective learning process. A combination of both systems methodologies (soft and hard sciences) and participatory action research can theoretically facilitate the integration of various disciplines and different types of knowledge. Such a combination is consistent with the definition of soft

systems methodology, and is referred to as collaborative modelling (Purnomo, Yasmi et al., 2003).

6.2.1.1. Diversity in Collaborative Group Composition

The types of disciplines and number of collaborators involved in the modelling process can determine the base of knowledge integrated into the model. Frequently, modelling has primarily been used by natural scientists as a means of capturing and predicting aspects of complex systems, usually within monodisciplinary boundaries (e.g. crop models, hydrological models). These kinds of “expert models” are inadequate to understand non-linear, complex and dynamic systems such as in the field of economics where several components of human-environment interactions are involved (Prell, Huback et al., 2007). Thus, economists have adopted modelling in different ways to integrate the product of several disciplines into an economic model.

However, these multidisciplinary models are usually implemented and used within the scientific arena. This kind of modelling approach is based on optimization and

‘factual knowledge’ that is insufficient for multi-level and multi-actor involvement, in particular rural development (Prell et al., 2007). The involvement of lay stakeholders throughout the model development process is proposed to avoid relying exclusively on knowledge derived from policy-makers and scientists (Purnomo et al., 2003). Therefore, the engagement of non-scientific stakeholders is needed if investigation of a RRM issue is targeted.

Participation of non-scientists in the production and use of scientific knowledge was introduced by social scientists who believe that science is socially constructed and should not be restricted solely to the scientific arena. Research on participation in science pay more attention to gaining support for decisions and enriching assessments with lay-knowledge and opinions (van Asselt, Mellors et al., 2001). Involving non-scientists in the research process is increasingly important in the analysis of complex issues. Such issues concern a tangled web of related problems problem), lie across or at the intersection of many disciplines (multi-disciplinary), and the underlying processes interact on various spatial and temporal scales (multi-scale) involving different stakeholders (multi-actors) (van Asselt and Rijkens-Klomp, 2002). However, indigenous knowledge alone is not sufficient

(Ostrom, Schroeder et al., 1993). Thus, integrating knowledge from different sources through collaborative processes can be a promising way for all relevant stakeholders to gain better understanding of the current situation of the system under study.

6.2.1.2. Collaborative Model: the Integration between Local and Scientific Knowledge

Knowledge integration that combines local and scientific knowledge is a challenging issue in scientific research (Neef, 2005). Local knowledge is primarily concerned with sustaining people’s livelihoods in harmony with nature. It is closed, non-systematic and bound to subjectivity, common sense or superstition. In contrast, science is open, systematic and objective, rigorous and analytical.

Because local knowledge is deeply embedded in the social, cultural and moral context, it is important to better understand this informative base of a society. Such information is needed to facilitate communication and decision-making resulting in the creation of an environment. This is conducive to collaborative processes that bridge the divide between local and scientific knowledge. Thus, in a collaborative modelling process, stakeholders play a proactive, central role in the design team, working together with modellers that can result in more practical designs and acceptable development (Schuler and Namioka, 1993). Through the collaborative and communicative platform, the knowledge-sharing activity is enhanced leading to shared representation developed among heterogeneous stakeholders (Ashby, 2003;

Narayan, 1996; Pahl-Wostl, 2006; Selener, 1997). The collaborative modelling process also supports capacity building (Fitzpatrick and Sinclair, 2003), helps to resolve conflicts and build consensus (Walkerden, 2006), and creates networking opportunity (Roux, Rogers et al., 2006).

6.2.2. Collaborative Modelling Phases

Collaborative modelling process is a continuous spiral of collective decision cycles generally consisting of five main phases: problem identification and structuring, model conceptualization, model implementation and validation, scenario exploration, and monitoring and evaluation of the model used (Daniell and Ferrand, 2006). The degree of stakeholder involvement, in the collaborative modelling process, is varied.

They can engage exclusively during the model design phase or simply be the end

users of the model; the model can also be co-designed with, and used by the stakeholders.

6.2.2.1. Model Design Phase Problem identification and formulation

This phase underlies the theory to elicit the individual perceptions to determine the nature of the situation where the problem is found, what elements are important to the problem, and how participants believe these elements to be related to one another over a variety of scales (Daniell et al., 2006). The aim is to create a shared understanding about the identity and extent of the problematic issues. It also allows a group of people to become acquainted with each other along with the development of problem structuring (Andersen and Richardson, 1997; Winz and Brierley, 2007).

In this phase, the involvement of a broad representation of stakeholder groups affected by the problem is essential. While the problem is collectively structured, the objectives to be achieved or situations to be avoided are also often determined. Once objectives are elucidated, potential processes, strategies or plans can be designed. The preliminary findings at this phase are useful for formalizing an initial conceptual model.

Model conceptualization

The model conceptualization is an analytical task that abstracts the system into a model described by elements of the system, their characteristics, and their interactions (Musselman, 1998). The representation of the users’ perceptions and conceptions in relation to the system where the users reside is reflected in the conceptual model (David, 2002). Maintaining stakeholder involvement and using information provided by participants during this stage can create trust in the modeller and the model-building process (Ford and Sterman, 1998). A conceptual model can be presented in various kinds of visualization ranging from simple and organized writing or diagrammatic flowcharts to computer-based applications. It can be used as a tool to facilitate effective and efficient communication between participants regarding a model’s structure, components and operations (van Asselt et al., 2001). Once the

initial conceptual model is formalized to sufficiently represent the system under study, it is developed to become an artefact aiming to further analyze the effect of potential actions may occur under defined problematic situations.

Model implementation and validation

This modelling stage captures the conceptual model using the constructs of a simulation language or system. Translation of the conceptual model into a programmed model constitutes the process of programming to build an executable simulation model. Additional information beyond the initial thinking is usually obtained during this model co-constructing activity. It includes re-examining the hypotheses used in the model, and refining the work of model development through verification and validation procedures. Three significant steps in the modelling procedure need to be defined in this section. They are calibration, verification and validation (Jorgensen et al., 2001). Calibration is an attempt to find the best accordance between computed and observed data by variation of some selected parameters (Hansen and Heckman, 1996). If not, the model is adjusted by modifying the values of certain parameters. Verification is a determinant of whether the computer implementation of the conceptual model is correct. It is a test of the internal logic of the model, and a subjective assessment of the behaviour of the model.

Once a model is operational after verification testing, the model needs to be checked for its validity and whether or not it is a good model of what it is supposed to represent. Validation is a determinant of whether the conceptual model can be substituted for the real system for the purpose of experimentation (Balci, 1998). An objective is to test how well the model outputs fit the data. Since the model is co-constructed with field collaborators, the model validation is often carried out in parallel with model implementation. Involving stakeholders can ensure that the content of the model is believable, its outcomes plausible, and that it sufficiently represents the problem being examined (Daniell et al., 2006). Successful projects depend on valid models, sound statistical analyses, and cogent reasoning (Musselman, 1998). It is important that model validity is accepted by stakeholders. Once the programmed model is completed and valid, it is ready to be used for running virtual experiments. The process of experimenting with the simulations is to evaluate system

behaviour, to analyze model sensitivity, to determine functional relations, and to train people involved in the modelling process (Balci, 1998).

6.2.2.2. Model Use Phase

This phase focuses on producing scenarios and management options by using the model. The scenarios to be analyzed should be collectively identified and explored by stakeholders. Possible solutions for desirable solutions are collectively analyzed through the outputs of simulations. However, no model can replace individual thought. Thus, the purpose of using models at this phase of the collaborative modelling process is to support the collective decision-making process, not to supplant it (Daniell et al., 2006). Neubert (2000) proposed two efforts that should be taken into account once carrying out scenario exploration. First, the discovery-oriented effort is to produce new knowledge about organizational or institutional innovation processes. Second, the literacy-oriented effort is to build individuals’ and communities’ capacities to handle their problems themselves. In reality, both discovery- and literacy-oriented efforts are often interlinked.

After exploring interesting scenarios, the synthesis of preferred actions drawn from desirable scenarios can be collectively assessed for the purposes of making a final decision and a collective agreement to implement an action plan. As stakeholders actively involved in the model design phase, their acceptance of and commitment to a model’s outcomes is high. Other positive outcomes can be increased independence, self-awareness, and empowerment of stakeholders to address local problems independently.

6.2.2.3. Monitoring and Evaluation

The monitoring and evaluation activity plays an important role in assessing the success of the collaborative modelling process. This activity is proposed to ensure that tasks are carried out according to the action plan. Any problem encountered can be treated on an ongoing basis through adaptive management. As an intervention measure, if the evaluation is carried out by the participants, they will be required to think about what is occurring in the process which can then potentially change their behaviour and have further impacts on the process and its outcomes (Daniell et al.,

2006). As an aid to the overall utility, outcomes and perceptions of the process can be used to determine how such a process can be improved.

6.2.3. Methods and Modelling Tools

Although the participants collaboratively working with modellers is important, methods and modelling tools used are also a determining factor in successful collaborative modelling. Group discussion is often a primary method in collaborative modelling. Other participatory methods (individual interviews, focus groups, etc.) can be used in the collaborative modelling process as well. Tools can be simple drawings or complicated computer simulations, or combination of both.

Methods and tools used in collaborative modelling aim to examine the system under study qualitatively and quantitatively. For qualitative methods, the cognitive mapping using software packages such as Decision Explorer and DANA, and problem structuring methods such as the Soft Systems Methodology and Strategic Choice Approach are usually acknowledged. There are many methods dealing with quantitative investigation raging from static representation for instance, spatial mapping through public participation GIS, to more dynamic models such as System Dynamics (STELLA and VENSIM), Multi-Agent Systems (CORMAS and REPAST), Multi-Criteria Analysis (PROMETHEE methods), and Probability and statistical methods, such as Bayesian Networks (Daniell et al., 2006). Six case studies with different methods and tools used are analyzed and compared in the next section.

6.3. Comparison of Six Case Studies in the Field of Collaborative Modelling for